Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
10401520 | Electric Power Systems Research | 2011 | 16 Pages |
Abstract
In this paper an improved method to denoise partial discharge (PD) signals is presented. The method is based on the wavelet transform (WT) and support vector machines (SVM) and is distinct from other WT-based denoising strategies in the sense that it exploits the high spatial correlations presented by PD wavelet decompositions as a way to identify and select the relevant coefficients. PD spatial correlations are characterized by WT modulus maxima propagation along decomposition levels (scales), which are a strong indicative of the their time-of-occurrence. Denoising is performed by identification and separation of PD-related maxima lines by an SVM pattern classifier. The results obtained confirm that this method has superior denoising capabilities when compared to other WT-based methods found in the literature for the processing of Gaussian and discrete spectral interferences. Moreover, its greatest advantages become clear when the interference has a pulsating or localized shape, situation in which traditional methods usually fail.
Related Topics
Physical Sciences and Engineering
Energy
Energy Engineering and Power Technology
Authors
Hilton de Oliveira Mota, Leonardo Chaves Dutra da Rocha, Thiago Cunha de Moura Salles, Flávio Henrique Vasconcelos,